How does it handle invalid chemical structures?

The skill includes robust error handling capabilities, such as an 'ignore_errors' flag, which allows the pipeline to continue processing large datasets even if some SMILES strings are invalid.

Is Molfeat compatible with standard ML libraries?

Yes, its transformers are scikit-learn compatible, making it easy to use with common frameworks like XGBoost, Random Forest, and scikit-learn Pipelines.

What types of molecular descriptors does Molfeat support?

Molfeat supports fingerprints (ECFP, MACCS), 2D/3D RDKit descriptors, Mordred descriptors, and pretrained deep learning embeddings like ChemBERTa and MolT5.

Can I use this for drug discovery tasks?

Yes, it is specifically designed for drug discovery applications such as virtual screening, property prediction (QSAR), and similarity searching.

Does it support 3D molecular featurization?

Yes, it includes 3D-specific descriptors like USR and USRCAT for capturing the three-dimensional shape and pharmacophoric features of molecules.

Molfeat Molecular Featurization

Name: Molfeat Molecular Featurization
Author: ricable

byricable

•

データサイエンスとML

Converts chemical structures into numerical representations for machine learning using over 100 specialized featurizers and pretrained models.

概要

Molfeat is a comprehensive chemical featurization toolset designed for molecular machine learning and drug discovery workflows. It provides a unified interface to over 100 featurizers, ranging from traditional fingerprints like ECFP and MACCS to state-of-the-art deep learning embeddings such as ChemBERTa and Graphormer. By enabling Claude to handle SMILES strings and RDKit molecules with scikit-learn compatible transformers, this skill simplifies the creation of QSAR models, virtual screening pipelines, and similarity search tools, featuring built-in parallelization and caching for high-performance data processing.

主な機能

1 GitHub stars
Support for SOTA models including ChemBERTa, ChemGPT, and Graphormer.
Scikit-learn compatible transformers for easy integration into ML pipelines.
Unified interface for 100+ molecular featurizers and pretrained embeddings.
High-performance batch processing with automatic CPU parallelization.
Integrated ModelStore for easy discovery and loading of chemical descriptors.

ユースケース

Quantifying molecular similarity for scaffold hopping and chemical space analysis.
Ranking compound libraries via virtual screening for drug discovery.
Predicting molecular properties and biological activity using QSAR modeling.

概要

主な機能

1 GitHub stars
Support for SOTA models including ChemBERTa, ChemGPT, and Graphormer.
Scikit-learn compatible transformers for easy integration into ML pipelines.
Unified interface for 100+ molecular featurizers and pretrained embeddings.
High-performance batch processing with automatic CPU parallelization.
Integrated ModelStore for easy discovery and loading of chemical descriptors.

ユースケース

Quantifying molecular similarity for scaffold hopping and chemical space analysis.
Ranking compound libraries via virtual screening for drug discovery.
Predicting molecular properties and biological activity using QSAR modeling.